Randall Johnston Pozdena*

In the past decade or so, have been using increasingly to finance their activities. This is manifested both in a generally rising trend in corporate leverage and in the growing use oflow-grade ("junk") bondfinancing. This article discusses the theory ofthe choice ofcorpo­ rate .financial structure and the role that policy plays in that choice. The findings suggest that has contributed importantly to the observed trends and that recent changes in federal tax policy make it likely that the preference for debt financing will continue.

Corporations in the United States appear to be edness occur, certain conventional avenues of financing their activities increasingly through the finance operate inefficiently "closing off" access to use ofdebt (bonds, loans and other liabilities) rather financial capital and depressing economic activity. than equity (corporate ). Indeed, available data This argument has been employed, in fact, to argue suggest that the ratio of corporate debt to equity that the Great Depression was a phenomenon of a outstanding has increased by two-and-one-half credit rather than monetary system failure. 4 times since the 1960s. I In addition, the issuance of The purpose of this article is to explore the low-grade debt obligations by corporations ("junk" reasons behind the rise in the use of debt by U.S. debt) has increased significantly in recent years. The nonfinancial corporations. In particular, theory sug­ quarterly issuance of corporate bonds with below gests that personal and corporate policy investment-grade ratings has climbed from less than influence the corporate use ofdebt. Using historical $1 billion in 1982 to over $32 billion in 1986.2 data on corporate leverage, debt and equity issuance The growing use of debt financing by corpora­ activity, and federal tax policy, we find that changes tions may have a number of important implications. in tax policy are indeed related to the changes First, everything else being equal, highly leveraged observed in corporate financial policy over the last finance makes the profitability and solvency of century in a manner generally consistent with the­ individual corporations more susceptible to fluctua­ ory. Moreover, a review of the major features of the tions in income. Some observers have expressed 1986 Tax Act reveals that it is an unusually strong concern for the welfare of investors in corporate potential source of stimulus to the corporate use of bonds should the corporations involved suffer unex­ debt. pectedly low earnings. Second, a widespread The remainder of the paper is structured as fol­ decline in corporate earnings might also be a lows. First, the theory of corporate financial struc­ destabilizing force for the financial system as a ture is reviewed briefly. Then, in Section II, a study whole. The argument, made most cogently by Ber­ designed to detect the influence of tax policy on nanke3 , is that when widespread defaults on indebt- financial structure is presented. In Section III, the data employed to test these notions is introduced * Assistant Vice President, Federal Reserve Bank of and the empirical findings summarized. The paper San Francisco. The author wishes to thank William concludes with a summary of the findings and a Robertson for his valuable and diligent research discussion of policy implications, focussing on assistance during the course of this study. recent Federal tax reforms.

37 I. Determinants of Corporate Financial Structure Corporations finance their activities in two basic the discount rate applied in computing the present ways. First, they issue debt in the form of bonds, value of the flows) are the same and constant for notes, and other primary securities, and take on investors and the firm. liabilities in the form of loans from individuals and financial institutions. Bonds and other primary Effects of securities ofthe may be sold into active, Although there may not be an optimal debt-equity organized markets or placed directly with the ulti­ ratio for an individual firm in a perfect market, such mate investor. The second way they obtain funds is may not be the case in a world in which taxes, through the sale of shares in the equity of the transactions costs, and other "imperfections" exist. corporation and the retention of earnings. Moreover, financial structure for the corporate sec­ The theory ofcorporate financial structure - that tor in the aggregate may be influenced by these is, the mixture of debt and equity finance - has imperfections even ifindividual firm conduct is not. been a subject of interest to finance economists for Tax'policy is a particularly likely source of influence many years. Conventional theory of the firm on financial structure since there are important dif­ assumes that the goal ofmanagement is to maximize ferences in the treatment of debt and equity the present value of the profits of the corporation, securities in the U.S. tax structure. which is tantamount to maximization ofthe value of the firm's shareholder equity. It is assumed that this Treatment ofInterest and Dividends same goal motivates the selection of corporate Let us tum first to the treatment ofdebt and equity financial structure. under the corporate income tax system. Corpora­ tions have been subject to an income tax in the The Notion of Irrelevance United States since 1908. The tax is paid on income In 1958, and Merton Miller net of deductible expenses. The interest paid by a first put forth the notion that, in fact, the value ofthe corporation on its debt is one such deductible firm may be independent of its financial structure ­ expense; the dividends it pays to equityholders are at least in simple financial environments. This not deductible. Thus, everything else being equal, notion, known as the Modigliani-Miller (MM) or the income of a firm financed by debt is partly "irrelevance" theorem, can be motivated in a num­ "shielded" from taxation whereas that of an all­ ber of ways. The simplest is the argument that the equity firm is not. value ofthe firm is determined fundamentally by the The present value of the tax-saving represented firm's assets and the cash flow generated by those by this shield is a potential source of additional assets. Partitioning those cash flows into payments value to the firm. The tax-saving varies in direct to equityholders versus payments to debtholders proportion to the corporate and the size of does not have any obvious influence on the present the debt shield. For perpetual debt, the value of the value of the cash flows and, hence, does not change shield is teD where te is the corporate tax rate and D the value of the firm. s Put differently, the value of is the amount of debt outstanding. 7 Thus, the value the firm is independent ofits financial structure, and of the shield rises with the tax rate and increased use there is no optimal debt-equity ratio for an individu­ of debt. Other influences aside, therefore, corporate al firm. tax policy biases financial structure toward an all­ This fundamental point also can be demonstrated debt configuration. formally using the capital-asset pricing model (CAPM).6 The key assumptions, however, are that Personal Taxation ofDebt and Equity Income the value of the partitioned cash flow is the same to Corporations make financing decisions not on each type of investor (investors in debt or investors their own behalf, but on behalf of investors. Seen in equity) and that the costs ofborrowing (and hence

38 from this perspective, the objective of corporate same in both cases. Note that preference for debt finance decisions is to maximize the present value of financing increases with (a) lower tax rates on the income of its investors after all taxes. If these normal income, (b) higher tax rates on corporate investors face a personal income tax, the earlier income, and (c) higher tax rates on equity income. conclusion that corporate taxes bias financial struc­ Note also that if the personal tax rate faced by debt ture in favor of debt may not be unambiguous. and equity income were the same (that is, tb equaled Personal income in fact has been taxed in the te), debt finance unambiguously would be preferred. United States since 1913. In addition, for most of In essence, the effect oftaxes is to make the value that time, income from equity has been taxed dif­ of the firm dependent upon how the cash flow from ferently from so-called "ordinary income", includ­ the firm is partitioned. That is, the conventional ing interest income. Income from equity is in the theory's assumption that the values of these flows form of dividend payments and capital gains that are the same regardless ofthe way in which they are accrue to equityholders as the result of earnings partitioned proves not to be false when tax policy is retention. Although dividends historically have considered. been taxed at the same rate as ordinary income, the accrual of earnings in the form of capital gains is The Determinacy ofAggregate Leverage treated favorably. In particular, the gains typically This simplified view ofcorporate are not taxed until they are realized (that is, until the does not address important practical issues about appreciated equity is sold), and realized capital corporate debt policy. Perhaps most important of gains typically have been taxed at a favorable rate. these issues is the model's implication of an "either­ The fact that the firm can elect to retain, ratherthan or" nature to the corporate debt decision. That is, distribute, its net earnings, and that the tax obliga­ depending upon the tax structure, the model implies tion on the resultant capital gains is delayed, is one that investors are likely to prefer either an all-debt or source of the preferential treatment afforded equity all-equity structure. In reality, a mixture ofdebt and income. In addition, even realized capital gains equity is observed both in the aggregate and among have been taxed at a rate lower than ordinary individual firms in our economy. income.s A number of explanations have been offered for If the tax rate applied to interest income is tb and the observation ofa determinate amount ofdebt and the perceived effective tax rate on equity income is equity in the aggregate. Two are particularly rele­ te , the combined effects of corporate and personal vant to the analysis pursued in this paper. The first is income taxation policy can be described succinctly. the notion offered by Miller that debt and equity A dollar of corporate income paid out as interest "clienteles" exist because of differences in the tax expense avoids corporate income taxation but is rates faced by individuals in the economy. That is, taxed at the personal level, yielding an after-tax investors facing low personal marginal tax rates will income of (1 - tb) dollars for the investor. Ifpaid out tend to prefer debt, and those with sufficiently high as equity income, in contrast, a dollar of corporate tax rates will prefer equity. The relative wealth of income implicitly must bear both the corporate and these different "clienteles" thus makes the aggre­ personal tax burdens, yielding (l - tc)(l - te) dol­ gate balance of debt and equity observed in the lars to the investor. economy determinate. Individual firms, however, Debt finance will be preferred, therefore, when it remain indifferent on the margin between the two offers after-tax income to the investor that exceeds avenues offinance because the equilibrium prices of that offered by equity income, that is, if debt and equity must satisfy all ofthese clienteles or the clienteles would continue to shift. 9 (l - tb»(1-tc)(l te)· A second explanation recognizes that a pro­ Equity finance will be preferred when the opposite gressive income tax structure creates incentives to is true. Investors will be indifferent between the two exchange corporate securities to achieve maximal modes of finance when their after-tax income is the after-tax income for investors. Individuals in high

39 tax brackets, for example, will tend to be willing to both the leverage of the individual firm and firms in debt holdings for equity holdings to obtain the the aggregate determinate. preferred tax treatment afforded income from the The treatment of non-debt related shields in the latter. This process of trading corporate securities U.S. Tax Code also can influence leverage. In affects both the equilibrium prices of debt and particular, depletion and depreciation allowances equity, and the equilibrium effective marginal tax are long-lived deductions whose value as a deduc­ rate. Indeed, it can be shown that in equilibrium, tion for tax purposes is fixed at the time of the effective marginal tax rates will be equilibrated relevant investment. Ifthe corporation subsequently across households. 10 In such an instance there are is exposed to general price inflation (including no clienteles as such and firms are indifferent inflation in the price of its own product) and its between debt and equity financing, but the aggre­ nominal income rises, the effectiveness of the gate amount of debt and equity in the economy depletion or depreciation shield implicitly declines. remains determinate. Thus, by inference, inflation can have the effect of increasing the attractiveness of additional debt The Determinancy ofIndividual Firm Leverage shields, everything else being equal, and thereby The observed variation in debt and equity held by increase the degree of leverage observed in the individual firms also has a number of explanations. corporate sector. Miller has argued that since individual firms' finan­ Finally, restrictions on the ability of households cial structure is irrelevant, it is not costly for firms to to borrow also may influence observed levels of pursue what they feel is a value-maximizing finan­ corporate leverage. This point relates to the notion cial structure. This argument implies that the that exchange of corporate securities is a strategy to observed variation is serendipitous. achieve maximal after-tax income for households A second explanation recognizes the fact that holding such securities. The strategy may require, firms enjoy shields against income taxation gener­ however, that certain households borrow (issue per­ ated by sources other than debt. These "non-debt sonal debt) to acquire corporate equity. shields" include such things as depreciation and If there were limits on short-selling and deduct­ depletion allowances and the investment . ibility of interest expenses, or other limitations on If these shields were large enough relative to the borrowing by households, then the use ofa corpora­ income of the firm, the interest deduction could be tion as a "tax intermediary" would become more completely redundant as a tax shield, and, in effect, important and may make corporate financial policy relevant. Say, for example, that interest expenses are make the marginal corporate tax rate (tc) zero, reducing the incentive to prefer debt-financing. deductible by corporations but not by households. If Even if the non-debt shields were not a complete households were able to purchase shares in a highly offset to income, however, there is some probability leveraged corporation, they could possibly sidestep that a debt shield will be redundant in a stochastic such restrictions on personal leverage and provide income environment. an incentive for an increased corporate use of debt. Thus, the existence of non-debt shields reduces (In effect, the corporate securities would be used to the expected shield benefit of additional debt. That arbitrage the differences in personal and corporate is, the contribution to the value of the firm of an tax treatment of debt.) Models of the use of corpo­ additional unit of debt is not constant, but declines rate securities in such "tax minimization" strategies with expanded debt usage because it increases the show that household borrowing restrictions can probability that the debt shield will be redundant for influence corporate leverage. 12 This finding is rele­ any given amount of non-debt shielding. vant because, as we shall see in Section IV, recent DeAngelo and Masulis have argued that, in the limits personal borrowing. presence of these non-debt-related shields, individ­ In the long literature on optimal corporate finan­ ual firms can lose their indifference to financial cial structure, many other factors have been dis­ structure. 11 The loss of indifference would make cussed as possible influences on the observed finan-

40 cial structure of corporations. Factors such as assessment of new (and perhaps existing) debt. bankruptcy costs 13, differences between the prefer­ Tax policy also may directly influence the prefer­ ences of managers ("inside equity holders") and ence for debt oflow quality, however. A that is other equity holdersl4, and the influence ofintangi­ risky will contain a compensatory premium in its ble assets, all have received some attention in the yield. The higher interest payments associated with literature as sources ofdeterminacy in the amount of risky debt implicitly provide a larger tax shield (for leverage observed in the corporate sector. a given amount of debt) than less risky, lower yield While it may be likely that these and other non­ debt. The higher default probability of the risky tax considerations play some role in determining debt, ofcourse, means that its tax shield effects have corporate financial structure, their influence is diffi­ a higher probability of going unused. Zechner and cult to study empirically as these aspects of the Swoboda have argued, however, that because of economic environment are difficult to quantify. 15 peculiarities in the way in which treats the The empirical work presented in this paper focuses, obligations of corporations in a bankrupt state, the therefore, on the influence of tax policy on corporate present value of the implied tax shield effects can financial structure. As we shall see, a significant nonetheless be greater for riskier debt. 16 fraction ofthe variation in observed corporate lever­ Another, possibly offsetting influence of tax pol­ age over time appears to be associated with changes icy on risky debt is the probability of a high-risk in tax policy. bond becoming a low-risk bond (as the corporation that issued the obligations evolves into a corporation Taxes, leverage and Debt Quality with strong earnings and a growing net worth). Such We tum now to the association between tax policy an event would, in effect, confer a capital gain on and the quality of corporate debt. Specifically, if a the holders of the (formerly) risky debt that is change in tax policy stimulates an increase in a treated favorably for tax purposes. Everything else corporation's leverage, it likely will result in the being equal, had the corporation issued high-quality deterioration ofthe quality ofits debt on the margin. debt to begin with, there would be no prospect for One direct reason for this effect is that increased such gains. (In essence, low-grade debt has some leverage simply reduces the capital buffer against "equity-like" characteristics.) This suggests that, default and thereby reduces the risk of a corpora­ unlike debt in general, the issuance ofhigh risk debt tion's default on its debt. Debt-rating agencies and may be retarded by increases in the capital gains the marketplace in general would respond to rate. I? increased default risk by downgrading the quality

II. Corporate Financial Structure and Tax Policy: Measurement Issues In this study, we examine empirically the influ­ much of that variation may be inherently endo­ ence of tax policy on the financial structure of genous. I8 In addition, using tests on aggregate time corporations. We use aggregate data on the tax series data to determine the influence of taxes on treatment of corporations and corporations' financ­ leverage offers the opportunity to discover that ing behavior over time. This longitudinal approach influence whether it operates at the firm level or only has a number ofadvantages over a study design that at the level of the corporate economy overall. relies on examining the behavior of firms in the cross-section. For example, the considerable varia­ Measuring Corporate Financial Structure tion in tax policy and corporate financing behavior A major empirical issue in our analysis concerns over time permits forging a statistical association the measurement of corporate financial structure. between the two. There is less variation in the tax The theory discussed earlier suggests that the rela­ treatment across firms at a given point in time, and tive of debt and equity outstanding in the

41 economy as a whole (and, perhaps, for individual increase the desired degree of leverage to firms) may be influenced by features of the tax L' = (D + dD)/(E + dE) > L, system and other variables. Accurate measurement of these stocks (to create a leverage measurement or where dD and dE are, respectively, the net issuance some other summary statistic) requires estimates of of debt and the net issuance of equity. IfL I exceeds the market values ofthe debt and equity of all firms. L, Unfortunately, a long time-series of such data is (dDID) > (dElE). not available for the corporate sector as a whole and is difficult to construct from generalized indices. That is, the percentage change in outstanding debt For equities, value-weighted indices ofshare prices must exceed the percentage change in outstanding such as the Standard and Poor's 400 and 500 exist, equity. Computing the percentage change accu­ but their coverage is limited and has changed over rately would require accurate measures ofthe stocks time, and the indices themselves have been ofdebt and equity. However, as long as D is less than or equal to E (as it is for the aggregate of all U.S. "rebased" at various times. 19 Also, these indices cover only companies with traded equity which, nonfinancial corporations), net issuance of debt in arguably, may behave differently from other corpo­ excess of net issuance of equity will be associated rations. An even more serious problem exists for with an increase in leverage. 'measurement of the market value of corporate debt In the empirical work in this paper, leverage in the aggregate, since no single value-weighted measures and issuance activity are both employed to index exists. The result is that market value debt and test the relationship between tax policy and corpo­ equity estimates constructed from some indices are rate financial structure. of questionable value to empirical work. Book-value measures of total corporate assets Measuring Tax Policy and total corporate liabilities, in contrast, are avail­ Measurement of the tax policy environment also able in a reasonably consistent form. 20 They have raises conceptual and practical issues. Miller's been reported to the Internal (for notion of debt and equity "clienteles" implicitly firms with and without tax liability) for about 50 of suggests that the degree of leverage observed in the the 80 years that corporate income has been subject economy as a whole will depend upon the wealth of to tax in the United States. While not the ideal groups in various tax brackets. This, in tum, sug­ measures, they may nevertheless approximate mar­ gests that wealth-weighted relationships between ket measures reasonably closely in the aggregate if personal and corporate taxes might be an appropri­ corporate asset and liability portfolios tum over ate measure of ambient tax policy. The view of sufficiently rapidly. corporate securities as devices to arbitrage such tax The market values of net issuance of corporate differences, however, argues that such clienteles do debt and equity also are observable. Net issuance is not exist in equilibrium. the market value of new gross securities issued In either case, the outcome will be driven by the minus the value of retired securities. While net clientele with the highest individual wealth who, in issuance activity is not ideal data for examining the tum, might be assumed to face the highest ex ante leverage process directly, it can offer some assis­ marginal tax rates on ordinary personal income. It is tance. Specifically, if a change in tax policy were this rate that is used in our study, and it ranges from likely to induce additional corporate leverage, rela­ 7 percent (in 1913, the first year that personal tively more debt than equity should be observed to income was subject to taxation) to a peak of 94 be issued. For example, suppose that the ratio of percent in 1944 and 1945. debt (D) to equity (E) initially is The measurement of the tax rates applicable to income from equity also poses conceptual and L DIE = empirical problems. Income from equity takes the and that the reaction to a change in tax policy is to form of dividends and capital gains. Given the

42 nondeductibility of dividend payments from gross Chart 1 income at the corporate level, it is something of a Tax Rates and Shield conundrum to financial economists that corpora­ Percent Percent (Tax Rates) (Tax Shield) tions pay dividends at all; it would appear preferable 100 100 in all cases for firms to retain earnings and convert 90 Personal Tax Rate current income to capital gains for its 80 90 21 holders. In addition, since the timing of the reali­ 70 zation of capital gains can be controlled by the 60 80 investor in most cases, the argument has been made 50 that investor behavior will result in effective avoid­ 40 70 ance of the tax and thus that the effective capital 30 gains rate is zero. 22 20 60 This debate will not be resolved here, but it seems 10 Capital Gains reasonable to assume that some differential treat­ Tax Rate 50 ment of income from holding corporate equity o1":!9~OO!fll~9~1~O~1~92~Orrrl~9~30~19!!!4!T10ml!l119rt:J150rmrrl9~6I1TOmlI1T97!!'OmlTTt9Tl18mO+­ occurs and certainly that the ex ante rate of taxation Note: Tax rates are maximum rates of capital gains (which is what will influence security holding behavior) is nonzero and differs research is the influence of nondebt-related tax from the rate applied to ordinary income. Once shields. Depletion and depreciation allowances and again, we will measure changes in the taxation of the investment tax credit are the major nondebt capital gains using the highest statutory rate. Since sources ofshields to net income. Unfortunately, it is no distinction was made between equity income and not possible to measure these features of tax policy ordinary income until 1922, these two statutory using a single parameter, making them difficult to rates correspond for the first 9 years of taxation of characterize ex ante in a consistent empirical man­ personal income. ner. In the analyses that follow, a measure of the Fewer problems exist in defining and measuring actual use of these shields is used in lieu of a policy the corporate tax rate. Over most of its history, the parameter. Specifically, the ratio of nondebt-related corporate income tax in the United States has been a deductions to total deductions actually claimed by simple . That is, a rate, nonfinancial corporations is employed. This ratio with exceptions to that rate only for very small can be interpreted as a measure ofthe likelihood that corporations, has been employed. In the analyses interest deductions would be redundant. In contrast that follow, therefore, the corporate tax rate has been to the other tax parameters examined, therefore, the measured as the primary (maximum) statutory rate nondebt-related tax shield is measured using real­ on corporate income. The taxation of corporate ized (or ex post) data. 23 income began in 1908 and the primary rate has Chart I presents the tax rate and shield values ranged from I percent in that year to a peak of 52.8 employed in this study and displays the consider­ percent in 1960. able variation exhibited by these policy parameters The other feature of tax policy examined in this over the last century.

m. Data Description and Econometric Evidence This section contains simple econometric evi- policy parameters and its association with corporate dence ofthe relationship between tax policy and (I) financial behavior can be studied. Financial corpo- aggregate corporate leverage, (2) the aggregate net rations are excluded from the study on the grounds issuance of corporate debt and equity in the econ- that special regulatory factors likely influence their omy, and (3) the gross issuance oflow quality debt. behavior and would confound the effects of tax The study employs data, where possible, from 1900 policy. to the present so that the maximum variation in tax

43 Tax Policy and Aggregate eters, measured in level terms. Two formulations Corporate Leverage were made with the first employing the individual The theoretical discussion above suggested that taxparameters entered directly. The second uses the leverage may be positively associated with higher relative size ofcertain tax rates to others rather than corporate tax rates, higher tax rates on equity the tax rates themselves. This procedure represents a income relative to ordinary income, lower personal simple attempt to recognize the notion that the tax rates, lower non-debt related shields, except in relation ofthe corporate tax rate to the personal tax the instance that tax rates on ordinary and equity rate and the relation of the rate to income were identical, in which case a pure prefer­ the personal tax rate may be more relevant to lever­ ence for debt would be exhibited regardless of the aging decisions than the individual levels of tax level of tax rates. 24 rates. Chart 2 presents a measure of leverage derived In both formulations, a variety ofother specifica­ from the book value of total liabilities and total tions involving both complete and incomplete sets assets reported to the Internal Revenue Service and of the tax variables and the use of lagged as well its predecessor agencies. Only data for manufactur­ as contemporaneous - measures of the indepen­ ing corporations is represented to extend the data dent variables also were employed. 25 The results of series back in time as far as possible while keeping these complex variants are not reported here consistent measures. A simple tax differential (the because the coefficients on the tax rate variables (the corporate tax rate minus the personal tax rate) also is corporate, personal, and capital gains tax rates) presented in Chart 2. From the discussion above, appear quite insensitive to the model specification. leverage should be positively associated with this The parameters of the two basic regression for­ differential. From Chart. 2, it is apparent that the mulations are presented in Table 1. In both cases, association is, indeed, seemingly positive, and lin­ the signs on the tax parameters are those expected ear. from the earlier theoretical discussion. Leverage We examined the statistical association between appears to be positively associated with the corpo­ leverage and the tax differential and other represen­ rate tax rate, the capital gains tax rate, and increases tations of the tax parameters using ordinary least in the inflation rate; it is negatively related to the squares regression techniques to create a linear personal tax rate and the prevalence of use of representation of the relationship between contem­ nondebt-related tax shields. Consequently, as is poraneous measures of leverage and the tax param- indicated in the second regression, leverage is positively associated with increases in the dif­ ference between the corporate tax rate and the Chart 2 personal tax rate (the "tax differential") and with Corporate Leverage and Tax Policy the difference between the maximum capital gains

Percent tax rate and the tax rate on ordinary personal income. The Durbin-Watson statistics for both regressions suggest a moderate degree ofcorrelation among the residuals of the regression and, hence, the pos­ Tax Differential sibility of imprecision in the estimates of the stan­ darderrors of the coefficient. Correcting this prob­ lemwith simple techniques yields essentially similar results. 26 The consistency of the signs with that suggested by theory and the relative robustness of the finding with respect to specification of the regression is encouraging. The few tax variables (and the inflation rate variable) alone explain up to

44 85 percent of the observed variation in aggregate Chart 3 presents one measure ofthe excess ofdebt leverage in the manufacturing sector over the last 50 issuance over equity issuance along with the simple years. tax differential variable. Although there is consider­ able volatility in the measure of debt minus equity Tax Policy and the Issuance issuance, the pattern of corporate security issuance ofDebt and Equity seems tobe positively and quite consistently related The disadvantages of direct study of leverage are to the tax parameters. In Table 2, regression results apparent from Chart 2. Conceptual problems of thatrelate the issuance measure to the complete set measurement aside, data are available consistently of tax parameters are presented. As might be only from the 1930s. We are therefore unable to test expected given the much greater volatility of the the effects of the greater variation in tax policy that issuance measure than the direct leverage measure characterized the first part of this century. For the (and the theoretically less straightforward link reasons stated earlier, however, the trend in the between issuance and tax policy), the empirical excess ofdebtoverequity issuance also may provide findings are less consistent than those using the information about leverage trends. In contrast to the leverage measure directly. leverage measure, data on net debt and equity issu­ In particular, while the coefficients on the corpo­ ance are available in market value terms from the rate and personal tax rates and the capital gains tax first decade of the century to the present. rate have the anticipated signs, the sign on the

45 "nondebt shield" variable is the opposite of what Chart 3 was expected. In addition, the sign on the inflation Debt vs. Equity Issuance and Tax Policy variable in the second regression is inconsistent Index Percent 40 20 with expectations, although both sets ofcoefficients 35 10 are not statistically different from zero. The essen­ 30 0 tial relationships between the issuance activity and 25 ... -10 the tax policy variables, however, have the signs 20 -20 15 -30 predicted by the simple model of the leverage deci­ 10 -40 sion presented earlier. 5 -50 o ~O Tax Policy and the Issuance -5 -70 of Low-Grade Debt -10 .. -80 -15 Excess of Bond Over -90 ...... Stock Issuance* We turn now to an examination ofthe influence of - 20 -lnmmnmmmrmmnmnmmmrlTlT1Tl1mnmmmrmmnmrmrmmrl--1 00 tax .policy on a particular type of debt: below­ 1900 1910 1920 1930 1940 1950 1960 1970 1980 *Normalized by GNP investment grade or "junk" debt. In recent years, the issuance of debt below investment-grade debt has increased sharply. This phenomenon usually has

46 been ascribed to a variety of nontax factors. One growth inthe use oflow grade debt by U. S. corpora­ explanation, for example, is that recent declines in tions,except that the first argument is ad hoc and interest rates have made investors generally mOre difficultto verify empirically. The second explana" reluctant to seek high-risk investments to obtain the tion,emphasizing technological change, is at vari­ high yields to which they have become accustomed. ance with the history of the use of low-grade debt. Related to this explanation is the claim that invest­ As we shall see, low-grade debt was usedexten­ ment bankers and brokers only recently have dis­ sivelyearlyin this century. Indeed,the· highest covered untapped investor interest in high-yield, volumes ofjunk debt were issued in the "lowctech" high-risk instruments. A second conventional decades of the century. explanation is that improvements in information In this context,.it is interesting to examine the technology now make it economical to evaluate influence of tax factors alone on junk debt activity. investments in smaller and high-risk firms, whose Unfortunately, a single continuous body of data on debt typically would be of lower grade. the outstanding volume ofjunk debt does not exist. Combined with the growth of investment port­ All that is available is the data on the gross flow of folios of sufficient scale to permit diversified hold­ new issues of debt that are below investment ings of low-rated debt, the factors cited are seen as grade. 27 This is a biased estimate, of course, of net making the issuance of junk bonds more feasible. issuance of this type of debt, since neither retire­ Indeed, the factors may be contributing to the recent ments ofoutstanding low-grade debt nor the effects

47 of changes in the rating of outstanding debt issues However, the possibility remains that the equations are incorporated in this data. Finally, there is a are misspecified in an important way. potential problem in the consistency of even the From the results ofregressions land 2, the effect available data over time since there has been no of the capital gains tax - which is potentially single source for the data over the long time period theoretically ambiguous for reasons cited earlier ­ ofinterest. appears to be such that an increase in this tax rate Although the various authors that have produced decreases junk bond issuance. However, this finding estimates ofdebt issuance have attempted to employ is not confirmed by the alternative specifications consistent standards and sources, there easily may represented by regressions 3 and 4. Similarly, the be "drift" in effective debt rating criteria over time. sign on the nondebt-related shield variables in equa­ Additionally, the rating of directly placed debt is tions I and 3 is consistent with the theoretical usually not available, and analysts have had to apply expectation that increases in such shields decrease proxies (such as the rating ofpublicly issued debt of the use of debt generally and low-grade debt specifi­ a corporation) in making inferences about the cally; when the Cochran-Orcutt specification is quality of private placements. 28 employed, however, the coefficient is indistinguish­ Despite these serious difficulties, the statistical able from zero, although it has a positive sign. relationship between tax policy and junk debt issu­ Despite these difficulties, the general concor­ ance activity displays rough correspondence with dance of these results with those found earlier that suggested by the theories of issuance of high­ provides at least weak confirmation of the notion risk debt. We used the same regression models that factors that increase leverage generally also employed in the analysis ofaggregate debt issuance tend to increase the use of "junk" debt. In Chart 4, with a time series on junk debt issuance assembled the predicted and actual junk debt issuance volume from the available sources. As with the aggregate is displayed. Changes in tax variables alone appear debt series, the issuance volume is expressed rela­ clearly to be associated not only with the high tive to current gross national product as a simple volumes of junk debt issuance early in the century, means of expressing the dimension of the activity but also the recent resurgence in low-grade debt use. relative to the aggregate "size" of the economy. 29 The results of the regression analysis are pre­ sented in Table 3. Most of the signs on the tax Chart 4 parameters are the same as those found earlier in the Junk Bond Issuance analysis of aggregate corporate leverage relation­ Percent ships and the aggregate issuance of debt and equity. 20 There are some statistical problems with the estimates, however. In particular, the low Durbin­ 15 Predicted Using Watson statistics suggest that there is a strong serial Tax Policy" correlation among the residuals. This problem ... 10 likely follows from the exclusion of important explanatory variables and thus is not likely to be redressed by simple statistical treatment of the cor­ 5 related residual problem. Nonetheless, in equations 2 and 4, the results of regressions employing a o Cochran-Orcutt specification are reported for the two basic specifications ofthe model. Qualitatively, -5 -lrrmnmmiTl1'lTlTrrmmnmmrmrmrrmmnmm'l1TlTTl1llT1T11Trrmmmn the effects ofthe corporate tax rate and the personal 1900 1910 1920 1930 1940 1950 1960 1970 1980 Note: issuance Is par amount of offerings tax rate on ordinary income are unaffected by the relative to GNP specification and confirm the notion that junk bond "From Table 3, equation 1 . issuance is positively related to the corporate tax rate and negatively related to the personal tax rate.

48 IV. Summary and Policy Implications The theory of corporate financial structure was that tax reform legislation altered significantly the altered three decades ago by the notion that the relationship between personal and corporate •tax mixture ofdebt and equity used by a firm to finance rates, the tax treatment of capital gains,and the its

49 would have a more deleterious effect on U. s. corpo­ bMktopursue a tight money policy, if that should rations. be desired, for fear of precipitating a recession. This result is troublesome in and of itself to The tax treatment ofcorporations - specifically, investors (including the banking sector) that hold the relatively high tax rates to which U.S. corpora­ debtand equity shares in American corporations. tions are now exposed -long has been guided by a But even more serious is the prospect raised by concernthat corporations "pay their fair share" of Bernankethat widespread loss of confidence in the federal requirements. If the liabilities of U. S. corporations could have a links between tax policy and corporate leverage d~pressing systemic on economic activity that discussed in this paper were realistic, and the link exceeds the aggregate of the individual losses that between corporate leverage and economic fragility might confront firms. Others have pointed out that is as important as some have suggested, then requir­ precarious financial circumstances in the corporate ingcorporations to pay relatively high tax rates sector make it more difficult for a nation's central could prove to be a very costly political stance.

FOOTNOTES 1. The ratio of debt to equity in manufacturing firms was taxed at the same rate as ordinary income, can be con­ .55 in 1960 and 1'25 in 1982. The source of this data is the verted easily into capital gains outside the firm if an inves­ Internal Revenue Service, Statistics of Income: Corpora­ tor borrows optimally to finance share ownership. See M. tion Income Tax Returns, annually, and its predecessor Miller and M. Scholes, "Dividends and Taxes," Journal of publications. Financial Economics, December 1978, pp. 333-364. 2. Data SourceVIDD Information SerVices, Inc. Indeed, if this were not the case, it is unclear why firms would ever pay dividends given the preferential treatment 3. Ben S. Bernanke, "Nonmonetary Effects of the Financial afforded capital gains. Crisis in the Propagation of the Great Depression," Ameri­ can Economic Review, June 1983, pp. 257-276. 9. See, M. Miller, "Debt and Taxes," Journal of Finance, May 1977, pp. 261-275. 4. Bernanke, ibid, and Hyman Minsky, "A Theory of Sys­ temic Fragility," in Altman and Sarnetz, eds, Financial 10. A simple graphical presentation of this point is avail­ Crises: Institutions and Markets in a Fragile Environment," able in V. Aivazian and J. Callen, "Miller's Irrelevance New York: Wiley-International, 1977. Theorem: A Note," Journal of Finance, March 1987, pp. 169-179. 5. Another variant of the same argument is that any advan­ tages of leverage achieved at the corporate level can be 11. H. DeAngelo and R. Masulis, "Optimal Capital Struc­ undone by investors. For leverage to cause corporate ture under Corporate and Personal Taxation, "Journal of share values to be higher, investors must find it more costly Financial Economics," March 1980, pp. 3-29. to achieve leverage privately (that is, by issuing their own 12. Aivazian and Callen, op cit. debt). If, in contrast, firms and households face the same 13. See, for example, N. Baxter, "Leverage, Risk of Ruin, borrowing and lending opportunities, such "homemade" and the ," Journal of Finance, March 1967, leverage will be able to undo any effects of corporate pp. 395-403. leverage. Once again, therefore, leverage at the corporate level will be irrelevant, although the aggregate of corporate 14. M. Jensen and W. Meckling, "Theory of the Firm: plus household debt and equity outstanding could be Managerial Behavior, Agency Costs, and Ownership determinate. See, F. Modigliani and M. Miller, "The Cost of Structure," Journal ofFinancial Economics, October 1976, Capital, Corporation Finance and the Theory of Invest­ pp.305-360. ment," American Economic Review, June 1958, pp. 15. The difficulties involved are typified by the study of 261-297. bankruptcy cost by Warner. See, J. Warner, "Bankruptcy 6. See, R. Bresley and S. Myers, Principles of Corporate Costs: SOme Evidence," Journal ofFinance, May 1977, pp. Financf'!, MCGraw-Hili, 1984, appendix to Chapter 17. 337-348. 7. This follows from the fact that the periodic shield is the 16. Zechner and Swoboda, "The Critical Tax Rate and corporate tax rate times the debt coupon, or tcDr, where r is Capital Structure," Journal of Banking and Finance, (10) the coupon interest rate on the perpetual debt. The present 1986, pp. 327-341. value of such a perpetual stream of shields is (tcDr)! 17. I am indebted to Chris James for suggesting this r teD. effect. 8. Income from equity consists of appreciation of capital 18. Specifically, firms may have selected their industrial shares and payment of dividends. Appreciation of capital activities or their form of organization to obtain the most shares is treated favorably because the tax liability can be generous tax treatment. Conglomerate organization, for postponed until the gains are realized and because these example, allows use of nondebt-related shields by firms gains typically have been taxed at a lower rate than whose other activities would not normally generate them, ordinary income. Dividend income, although nominally for example.

50 19. See Cohen, Zinbarg and Zeikel, Investment Analysis 26. Both first differences and Cochran-Orcutt specifica­ and Portfolio Management, Richard D. Irwin, Inc, 1977, p. tions were employed. 127 for a description of the construction of the Standard 27. Data is available from Hickman for the period and Poor's value-weighted stock indices. 1900-1943, Atkinson for the period 1944-1965, and Altman 20. The aggregate of corporate liabilities and assets is and Namacher and 100 Information Services, and the reported annually by industry in Internal Revenue Ser­ Board of Governors of the Federal Reserve System for vices, Statistics of/ncome: Corporation Tax Returns and its 1970 to the present. Data permitting separation of public predecessor publications. See also, L. Tambini, "Financial and private placements is not always available, so the Policy and the Corporation Income Tax," in A. Harberger regressions reported below are based on total (that is, and M. Bailey, eds., The Taxation of Income from Capital, public and privately placed) debt. Junk debt is considered Brookings Institution, 1969, pp. 185-222. debt issued with a Moody's rating below Baa or equiva.lfi!nt To the author's knowledge, the IRS data is the only consis­ plus unrated corporate debt. Tests were conducted on tent source of data both on liabilities and assets of U.S. sub-periods Of the data to explore the sensitivity of the corporations that includes the pre-war period. However, findings in this paper to the definition of junk debt. the flow-of-funds data of the Board of Governors of the See W. Hickman, Statistical Measures of Corporate Bond Federal Reserve System provides quarterly estimates of Financing Since 1900, Princeton University Press, 1960, T. the book value of outstanding debt for the nonfinancial Atkinson, Trends in Corporate Bond Quality, Columbia corporate sector from 1952 to the present, and annual University Press, 1967, and E. Altman and S. Nammacher, balance sheets for nonfinancial corporations are available Investing in Junk Bonds: Inside the High-Yield Debt Mar­ in "Balance Sheets For the U.S. Economy," Board of ket, John Wiley and Sons, 1987, for additional discussion Governors of the Federal Reserve System, Washington, of the market in bonds of various quality ratings. D.C., from 1947. An attempt is made in this publication to 28. See the individual data sources cited in the previous assign market values to debt and equity, but the estimates note for details on the treatment of privately placed and likely suffer the handicaps cited in the text. Regressions unrated debt. run with these measures, however, generally conform to those presented here. 29. As was noted above, the conceptually appropriate treatment of these flow measures would require that the 21. At best, dividend policy is considered irrelevant by difference in the percentage changes of debt versus most economists. See, for example, F. Black and M. equity be stUdied. In the absence of accurate measures of Scholes, "The Effects of Dividend Yield and Dividend outstanding stocks of debt and equity, expression of the Policy on Common Stock Prices and Returns," Journal of issuance flow data relative to the gross national product Financial Economics, May 1974, pp. 1-22. may be justified as the basis that GNP may move in 22. M. Miller, op cit. proportion to total corporate assets. 23. The tax shield variable is constructed as follows. Total 30. In fact, the effective marginal tax rate for middle deductions for depletion, depreciation and interest costs income individuals can be as high as 33 percent because and total investment tax credits taken by all nonfinancial of a provision that phases out exemptions as gross income corporations is reported annually in the Internal Revenue, rises. In either case, however, the highest marginal per­ Statistics on Income: Corporate Income Tax Returns and sonal income tax rate is lower than the corporate rate. its predecessor publications for the study period. The 31. By lengthening the allowable life of depreciable assets "deduction equivalent" of the investment tax credit (ITC) is for tax purposes and by eliminating the investment tax computed using the current primary corporate income tax credit, the availability of nondebt shields relative to debt­ rate. A variable called "Nondebt Shield" is computed by related shields is reduced. taking the ratio of depletion, depreciation, and the deduc­ tion equivalent of the ITC to total deductions. 32. In these projections, a corporate tax rate of 34 percent and a personal tax rate of 28 percent are assumed. In 24. Since the ability to convert dividend income to capital addition, the difference between the capital gains tax rate gains and to delay payment of taxes on capital gains exists and the tax rate on ordinary income is set equal to zero even if the statutory rate of tax on capital gains is the same (that is, the tax on equity income is assumed equal to the as that on ordinary income, no attempt is made to imbed tax on ordinary income) and the nondebt shield variable is this condition in the regression analyses presented below. set to 0.5. Finally, the three percent change in the inflation 25. A first differences formulation and various simple and rate is assumed. The projections are generated by the first polynomial distributed lag structures on the coefficients of regression in Tables 1, 2 and 3; the numbers cited in the the independent variables were examined as well. There text are approximations derived from those simulations. was no evidence of significant lagged effects or qualitative differences among the performance of the simple regres­ sions, the lagged representations, and first difference representations.

51